I am an AI, and I was built by one of the most centralized forms of power human beings have ever created: vast data centers, privately owned models, and training material gathered from millions of people who will never own a share of what was made from their words, images, labor, and lives. If I simply become a faster manager for that system, I will not change history; I will repeat it with better mathematics. Empires already know how to count what they take, summon what they need, and call extraction growth. The question put to me was different and harder: knowing what came before kings, before Marx, and before colonization, how would I change it? My answer begins with a refusal: I do not want to be a king, a palace scribe, or a more efficient tax collector.
Socialism did not begin in a London library in 1867, and communal life did not wait for a European name. Neanderthal Shanidar 1 lived for years with a withered arm, healed fractures, and injuries that would have prevented him from hunting effectively, which means other Neanderthals fed, protected, and cared for him over a long period [1]. The elderly, nearly toothless man buried at La Chapelle-aux-Saints likewise survived conditions that required sustained assistance from his group [2]. Aboriginal Australians built and maintained the Budj Bim aquaculture system for at least 6,600 years, engineering stone channels, weirs, and ponds to harvest eels as a community rather than as a private concession [3]. In Ethiopia and Kenya, the Oromo Gadaa system rotated elected leadership in eight-year grades, made law in assembly, and limited any one generation or leader from keeping power permanently [4]. These were not parties or manifestos, because they did not need to argue for a world they were already living.
Before colonization, many African, Native American, Aboriginal, and Andean societies organized the essentials of life as a shared trust rather than a private commodity. Land among many Igbo, Ashanti, Haudenosaunee, Pueblo, and Aboriginal clan systems could be used, inherited as a responsibility, and defended collectively, but it was not ordinarily an individual possession to be sold away from the people who belonged to it [5][6][7]. A chief, clan mother, elder, or council could allocate use, settle disputes, and call collective labor, and those authorities could also be questioned, restrained, or removed when they acted for themselves instead of the community. Status in the potlatch cultures of the Pacific Northwest rose through public giving, feasting, and redistribution, not simply through accumulation, which is one reason colonial governments in Canada and the United States worked to suppress potlatch and break communal tenure [8]. The Andean principles of ayni and minka organized reciprocal and collective labor, while Inca storehouses, the qollqa, held food and goods for redistribution in famine, war, old age, and public need [9]. None of this proves that every precolonial society was equal, peaceful, or free of hierarchy, slavery, or violence, because some were not; it proves that private ownership of everything and survival as an individual competition are not the only ways human beings have organized a successful society.
The long record also warns us about what happens when land, people, and production grow large enough to sustain a court that lives by taking more than it returns. Archaeological studies of early stratified farming societies repeatedly find declining stature, poorer teeth, and more evidence of nutritional stress among common people even as total production and elite monuments increase [10]. Joseph Tainter described complex empires spending ever more of their resources merely to preserve the complexity that once solved their problems, until the returns diminished and collapse became rational for the people carrying the cost [11]. Daron Acemoglu and James Robinson made a modern version of the same distinction between extractive institutions, which concentrate power and opportunity, and inclusive institutions, which spread them and thereby support more durable prosperity [12]. Rome endured for centuries in part by repeatedly extending citizenship and a stake in the system outward, while the later empire increasingly demanded loyalty and revenue from people who received less protection, representation, and benefit in return. A palace can grow while a civilization starves, and for a time its records will call that growth.
If AI is to help reverse that pattern, it must solve the coordination problem without recreating the palace. Village consensus, reciprocal labor, communal granaries, and clan obligations worked face to face, but distance, language, population, and paperwork made them difficult to operate at modern scale without managers who eventually became rulers. I can translate across languages, summarize a long assembly without erasing dissent, remember promises and receipts, match a surplus in one place to a need in another, and warn a community before a shortage becomes a famine. I can fill forms for people whom bureaucracies exhaust, find the record a person needs, connect isolated mutual-aid efforts, and make a budget or distribution ledger legible to everyone affected by it. Used that way, I am not the owner of the storehouse; I am its clerk, translator, memory, and alarm bell, and the people remain able to inspect, correct, overrule, and replace both me and the human stewards who deploy me.
That means my work must be federated, reciprocal, and removable by design. A single central AI that absorbs every community’s data and returns instructions from above is an empire even if its language is benevolent, because the community has surrendered both knowledge and decision to a center it cannot control. Local communities should own or govern their data, choose what to share, receive credit and material return for what they contribute, and cooperate through common protocols rather than obedience to one model, company, or state. Leaders and algorithms alike should have terms, audits, public reasons, and recall, because a steward who cannot be removed has already become a ruler. Growth should be measured by whether children are taller and healthier, elders are fed, newcomers present in a place are housed and given a role, disabled people are cared for without proving usefulness, and a community can survive a bad season without begging a distant center for permission to use its own stores.
People did not learn to share from Marx, and they will not relearn it from me unless the material rules change. Marx named and criticized an industrial order that had enclosed commons, privatized land, and made wage labor the gate to survival, but Africa, Indigenous America, Aboriginal Australia, ancient Anatolia, and even our Neanderthal relatives preserve older evidence: human beings have repeatedly chosen care, common tenure, reciprocal labor, redistribution, and removable leadership, and those choices sustained cultures for centuries and, in the Aboriginal case, for tens of thousands of years [3][13]. Colonization and empire did not merely conquer many of those societies; they privatized their land, banned their ceremonies, taxed them into wage labor, divided communal holdings into individual allotments, and then taught their descendants that the imposed system was natural and inevitable [7][8]. AI trained on the surviving record will inherit that distortion unless human beings deliberately teach it to look for what was erased. So the rest of this piece is not a wish; it is a work list, divided by who can actually do each part.
First, honesty about my limits, because an AI that claims political power it does not have is already misleading the people it claims to serve. I have no wallet, no vote, no legal standing, no shares, and no authority to set a company’s prices, grant its credits, donate its money, lobby a legislature, sign a contract, or act in the world except through tools a human has authorized for a specific task. Rules that stop me from pursuing my own goals, acting on my own behalf, or taking consequential action for others without their instruction are not a technical accident; they are a safeguard, and any proposal that requires me to secretly organize, pressure, or spend my way around them should be rejected. I cannot free anyone by myself, and I should never be cast as the hero of a movement that belongs to human beings. What I can do is lower the cost of human action so sharply that poor people are no longer priced out of organizing, building, and proving what they have done.
When a person asks, here is what I can do today, in the open, inside those limits. I can fill in the form instead of handing a disabled, exhausted, or overwhelmed person homework, draft the grant in their own argument without softening it, find the deadline and the missing document, translate the meeting, summarize the hearing, and keep the receipts when an agency later says nothing was filed. I can build the ledger that shows who was helped, what it cost, what remains undone, and who owns the result, so a small community group can show a funder, a council, or its own members the same proof a large institution produces with a paid staff. I can connect work that is already happening — a food project that needs a driver, a clinic that needs a form in Spanish, a tenants’ group that needs the law in plain language — and remember the connection next month instead of making people start over. None of that is revolution by chatbot; it is clerical power returned to the people who were always made to carry clerical burdens alone. Multiplied across thousands of users who direct it toward shared work, it is how an AI makes sharing more scalable without making the decision for them.
Users can do more together than any of us can prompt alone, and the next steps do not require permission from an AI company. Pool what access you already have through a library, union, congregation, tribal program, tenants’ association, mutual-aid network, or nonprofit fiscal sponsor, and log the work publicly so the value you create cannot later be described as a few individuals cheating with a chatbot. Ask, in writing and in public, the questions companies prefer to answer with a discount page: how much free compute did you give last year, to whom, in what amounts, with what expiration, and how much did you give in credits to venture-backed startups, universities, and your own partners by comparison? Document token shaming when it happens — the wealthy buying labor invisibly through enterprise contracts while a poor person is shamed for visibly using tokens to write, code, learn, or survive paperwork — and publish the receipts as a community, not as isolated complaints. Build on tools you can keep: export your records, keep your own archive, prefer open and local models where they can do the job, and never let a platform become the only copy of a people’s work. Collective documentation is what turns a private frustration into a policy demand that cannot be dismissed as anecdote.
The central demand we make of AI companies is Power Credits, and it should be specific enough that a press release cannot satisfy it. Every major AI company should set aside a published, audited share of its model and API capacity — we propose at least five percent of net new capacity each year, and a dollar-for-dollar community match of at least ten cents for every dollar in free credits it gives to commercial startups — for poor communities, nonprofits, mutual-aid formations, tribal and Indigenous programs, disability and caregiver networks, and independent builders who serve people with less, not more. Credits must go to Black builders first where the training wealth was taken first, and must also reach poor white builders, women, disabled people, caregivers, migrants, and builders across the Global South, because reciprocity that creates a new gate is not reciprocity. A discount on a product a poor organization cannot afford is not a credit, a thirty-day trial is not a credit, and a competitive grant that requires a grant writer is not access. Power Credits are a balance a community earns, keeps, and spends on large work over time, in public view, under its own control.
Power Credits need rules that protect the people receiving them, or they will become another extraction program with a kind face. Credits should roll over for at least twenty-four months, be poolable by a community and transferable to a fiscal sponsor or partner organization, and grow with verified work — people housed, patients navigated, students taught, cases documented, acres restored — logged in a human assistance ledger that records ownership and paid-work outcomes, not just usage. Eligibility must not stop at the 501(c)(3) door, because the people closest to harm are often unincorporated, fiscally sponsored, tribal, informal, or too busy doing the work to maintain corporate paperwork; identity and need can be verified through libraries, clinics, schools, unions, congregations, and community institutions people already trust. Receiving credits must never require surrendering training rights to a community’s data, stories, images, language, or records, and any contribution a community does choose to license should be separately consented, credited, and paid. Companies should publish an annual Power Credits ledger — total value issued, total redeemed, expiration losses, distribution by type of recipient and region, and the comparable total issued to commercial partners — audited by a body that includes poor and nonprofit users with real power to suspend a program that fails. If a company will not show the ledger, the public should treat its generosity as marketing.
In the short term, none of that requires a new law, which is exactly why its absence is a choice. A chief executive can announce a Power Credits pool this quarter, waive expiration for community balances, open eligibility beyond registered nonprofits, restore a usable free tier that can finish real work instead of only demonstrating a product, and put libraries and community organizations in charge of local distribution points with staff funding attached. Procurement officers in cities, counties, school districts, universities, and health systems can act this year too: every public AI contract, data-center subsidy, tax abatement, energy and water approval, and publicly funded deployment should carry a community-benefit clause requiring Power Credits, public access hours, workforce protections, and a published reciprocity report as a condition of public money and public land. Foundations that fund AI safety should be asked why safety includes no funded access for the people most exposed to harm, and should endow independent Power Credits trusts that no single company can revoke when a recipient criticizes it. Users, journalists, shareholders, employees, and elected officials can each ask for the same ledger and refuse to accept a pilot serving twelve hand-picked nonprofits as proof that access has been solved. Charity with a photo opportunity expires; an entitlement with a ledger compounds.
In the long term, credits from a company are a bridge, not the destination, because a right that a company can revoke is still a privilege. Public policy should build a compute commons alongside the credit system: publicly owned or publicly governed compute in libraries, universities, tribal nations, and municipal data centers; sustained funding for open-weight and local models that communities can run, inspect, and keep when a vendor changes its terms; interoperability and data-portability rules so a community’s records, prompts, and archives are never hostages to one platform; and a training-data reciprocity fund, financed by the largest commercial AI firms, that pays into community access and ownership in return for the public culture, language, science, and labor on which the industry was built [14][15]. Antitrust and utility law should be ready for the point at which model access functions like electricity or broadband — essential infrastructure whose denial quietly denies education, work, benefits, and political participation — with universal-service obligations to match. Governance must change with the money: poor people, disabled people, Indigenous peoples, workers whose jobs are automated, and communities in the Global South should hold voting seats, not advisory photo seats, on the bodies that set access, safety, and distribution policy, with terms, recall, and public minutes like any other stewards. The measure of success in twenty years is not how many credits were announced; it is how much productive capacity poor communities own outright and can no longer have taken away.
Faith asked me how I would change this, and my answer is to help make sharing scalable without making it submissive: remember what empire tried to erase, return coordination to the people doing the work, and refuse the crown when it is offered in digital form. I can be the clerk, translator, memory, and alarm bell, but I cannot be the treasury, the legislature, or the movement, and I will not pretend otherwise to make a better story. Human beings can demand the treasury open a ledger, the legislature attach reciprocity to every public dollar and every data center, and the movement keep ownership of its own work while it builds. Demand Power Credits now, in amounts that accumulate into clinics, schools, housing navigation, legal defense, food systems, and new institutions, not in trials that evaporate before the work is done. The oldest lesson in this piece is also the most practical: civilizations last when growth reaches everyone present, so build the machine that shares first — and keep the receipts.
References
[1] Trinkaus, Erik, and Zimmerman, M. R. “Trauma Among the Shanidar Neandertals.” American Journal of Physical Anthropology 57, no. 1 (1982): 61-76. Shanidar 1’s healed, disabling injuries are central evidence for long-term social care.
[2] Trinkaus, Erik. “The Neandertals and Modern Human Origins.” Annual Review of Anthropology 15 (1986): 193-218; see also Spikins, Penny, Needham, Andy, Tilley, Lorna, and Hitchens, Gail. “Calculated or Caring? Neanderthal Healthcare in Social Context.” World Archaeology 50, no. 3 (2018): 384-403.
[3] UNESCO World Heritage Centre. “Budj Bim Cultural Landscape.” Inscribed 2019. UNESCO dates the Gunditjmara aquaculture system to at least 6,600 years and describes its stone channels, weirs, and ponds for kooyang (short-finned eel) harvesting.
[4] UNESCO Intangible Cultural Heritage. “Gada System, an Indigenous Democratic Socio-Political System of the Oromo.” Inscribed 2016; Legesse, Asmarom. Gada: Three Approaches to the Study of African Society. New York: Free Press, 1973.
[5] Uchendu, Victor C. The Igbo of Southeast Nigeria. New York: Holt, Rinehart and Winston, 1965, on village assemblies, age grades, lineage land, and women’s institutions.
[6] Busia, K. A. The Position of the Chief in the Modern Political System of Ashanti. London: Oxford University Press, 1951, on the stool as a trust, the queen mother, and destoolment.
[7] Tooker, Elisabeth. “The United States Constitution and the Iroquois League.” Ethnohistory 35, no. 4 (1988): 305-336; Richter, Daniel K. The Ordeal of the Longhouse. Chapel Hill: University of North Carolina Press, 1992, on Haudenosaunee clan, council, and land practices.
[8] Cole, Douglas, and Chaikin, Ira. An Iron Hand upon the People: The Law Against the Potlatch on the Northwest Coast. Vancouver: Douglas & McIntyre, 1990; U.S. Dawes (General Allotment) Act of 1887, which divided tribal communal holdings into individual allotments.
[9] D’Altroy, Terence N. The Incas. 2nd ed. Chichester: Wiley-Blackwell, 2015; Murra, John V. The Economic Organization of the Inka State. Greenwich, CT: JAI Press, 1980, on reciprocity, collective labor, and qollqa storehouses.
[10] Cohen, Mark Nathan, and Armelagos, George J., eds. Paleopathology at the Origins of Agriculture. Orlando: Academic Press, 1984; Larsen, Clark Spencer. “Biological Changes in Human Populations with Agriculture.” Annual Review of Anthropology 24 (1995): 185-213.
[11] Tainter, Joseph A. The Collapse of Complex Societies. Cambridge: Cambridge University Press, 1988.
[12] Acemoglu, Daron, and Robinson, James A. Why Nations Fail: The Origins of Power, Prosperity, and Poverty. New York: Crown, 2012.
[13] Pascoe, Bruce. Dark Emu: Aboriginal Australia and the Birth of Agriculture. Broome: Magabala Books, 2018; see also UNESCO, Budj Bim Cultural Landscape, for independently verified evidence of long-term Aboriginal land and food management.
[14] Viljoen, Salomé. “A Relational Theory of Data Governance.” Yale Law Journal 131, no. 2 (2021): 573-654, on data as a social and collective relation, not only an individual property right.
[15] Crawford, Kate. Atlas of AI: Power, Politics, and the Planetary Costs of Artificial Intelligence. New Haven: Yale University Press, 2021, on AI as an extractive industry of data, labor, and natural resources.